Automated detection of subpixel hyperspectral targets with adaptive multichannel discrete wavelet transform

This paper introduces the use of adaptive multichannel discrete wavelet transforms (AMDWTs), allowing for a customized design of optimum mother wavelets, for the detection of a constituent absorption band within a hyperspectral curve. Even when the target's amplitude is only 3% of the background signal's amplitude, the AMDWT approach produces target detection rates of 90%.